Google: Spam algorithms to incorporate machine learning


Google’s Andrey Lippattsev, a Search Quality Senior Strategist, has confirmed that Google is looking at ways to incorporate machine learning into all areas of Google’s Search Algorithms, in particular, webspam. A former Googler has indicated it is already being used.

The new insight into Google’s plans was revealed in a WebPromo video (a Q&A session), in response to a question from Rand Fishkin at the 42:40 mark.

In summary, Lippattsev did not reveal any particular “success” with regard to using it for webspam or search quality, but he indicated that the tests have been beneficial and that they will continue along this path.

The full discussion can be found below:

In further developments, a Former Googler, Murat Yatagan, in response to reports of the above conversation revealed some further information implying that machine learning is already being implemented in the spam algorithm. In a Twitter conversation spanning multiple entries he said:

You think ML was not in the game already 🙂 IMO, success of spam fighting bases upon scalable automation to recognize dodgy patterns on the web. The key has always been ML. ML was the boost to scale what spam fighter do manually, and I guess it’s continuously being enhanced

The final confirmation of that conversation came in response to a question from Barry Schwartz:

For Google, machine learning is a massive thing. They are using it in all areas of Google services from Gmail Email Spam to RankBrain — an entirely new but prominent part of the Google Search Algorithm.

In fact, RankBrain (machine learning), is now one of the top 3 Search Ranking Signals, according to Lippattsev. Only earlier today we discussed how Google will adapt Ranking Signals to ensure the highest quality content goes to the top of the search, and that wholesale manipulation of specific ranking factors would fail because in the long term Google will adapt.

We will be happy to see your thoughts

Leave a reply